static void findOverlap(const Graph& g,
ContigID refID, bool rc,
const ContigNode& pair,
const DistanceEst& est,
OverlapGraph& out)
{
if (refID == pair.id()
|| (est.distance >= 0 && !opt::scaffold))
return;
ContigNode ref(refID, false);
const ContigNode& t = rc ? pair : ref;
const ContigNode& h = rc ? ref : pair;
if (out_degree(t, g) > 0 || in_degree(h, g) > 0
|| edge(t, h, out).second)
return;
bool mask = false;
unsigned overlap
= est.distance - (int)allowedError(est.stdDev) <= 0
? findOverlap(g, t, h, mask) : 0;
if (mask && !opt::mask)
return;
if (overlap > 0 || opt::scaffold)
add_edge(t, h, Overlap(est, overlap, mask), out);
}
/** Return a path, complemented if necessary. */
static ContigPath getPath(const Paths& paths, const ContigNode& u)
{
if (isPath(u)) {
unsigned i = u.id() - Vertex::s_offset;
return u.sense() ? reverseComplement(paths[i]) : paths[i];
} else
return ContigPath(1, u);
}
/** Return the sequence of the specified contig node. The sequence
* may be ambiguous or reverse complemented.
*/
static Sequence sequence(const Contigs& contigs, const ContigNode& id)
{
if (id.ambiguous()) {
string s(id.ambiguousSequence());
if (s.length() < opt::k)
transform(s.begin(), s.end(), s.begin(), ::tolower);
return string(opt::k - 1, 'N') + s;
} else {
const Sequence& seq = contigs[id.id()].seq;
return id.sense() ? reverseComplement(seq) : seq;
}
}
/** Return the sequence of the specified contig node. The sequence
* may be ambiguous or reverse complemented.
*/
static const Sequence getSequence(ContigNode id)
{
if (id.ambiguous()) {
string s(id.ambiguousSequence());
if (s.length() < opt::k)
transform(s.begin(), s.end(), s.begin(), ::tolower);
return string(opt::k - 1, 'N') + s;
} else {
string seq(g_contigs[id.id()]);
return id.sense() ? reverseComplement(seq) : seq;
}
}
/** Return the sequence of the specified contig. */
static string sequence(const ContigNode& id)
{
const string& seq = g_contigs[id.id()];
return id.sense() ? reverseComplement(seq) : seq;
}
/** Return whether this vertex is a path or a contig. */
static bool isPath(const ContigNode& u)
{
return u.id() >= Vertex::s_offset;
}
/** Identify paths subsumed by the specified path.
* @param overlaps [out] paths that are found to overlap
* @return the ID of the subsuming path
*/
static ContigID identifySubsumedPaths(const Lengths& lengths,
ContigPathMap::const_iterator path1It,
ContigPathMap& paths,
set<ContigID>& out,
set<ContigID>& overlaps)
{
ostringstream vout;
out.clear();
ContigID id(path1It->first);
const ContigPath& path = path1It->second;
if (gDebugPrint)
vout << get(g_contigNames, ContigNode(id, false))
<< '\t' << path << '\n';
for (ContigPath::const_iterator it = path.begin();
it != path.end(); ++it) {
ContigNode pivot = *it;
if (pivot.ambiguous() || pivot.id() == id)
continue;
ContigPathMap::iterator path2It
= paths.find(pivot.contigIndex());
if (path2It == paths.end())
continue;
ContigPath path2 = path2It->second;
if (pivot.sense())
reverseComplement(path2.begin(), path2.end());
ContigPath consensus = align(lengths, path, path2, pivot);
if (consensus.empty())
continue;
if (equalIgnoreAmbiguos(consensus, path)) {
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n';
out.insert(path2It->first);
} else if (equalIgnoreAmbiguos(consensus, path2)) {
// This path is larger. Use it as the seed.
return identifySubsumedPaths(lengths, path2It, paths, out,
overlaps);
} else if (isCycle(lengths, consensus)) {
// The consensus path is a cycle.
bool isCyclePath1 = isCycle(lengths, path);
bool isCyclePath2 = isCycle(lengths, path2);
if (!isCyclePath1 && !isCyclePath2) {
// Neither path is a cycle.
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "ignored\t" << consensus << '\n';
overlaps.insert(id);
overlaps.insert(path2It->first);
} else {
// At least one path is a cycle.
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "cycle\t" << consensus << '\n';
if (isCyclePath1 && isCyclePath2)
out.insert(path2It->first);
else if (!isCyclePath1)
overlaps.insert(id);
else if (!isCyclePath2)
overlaps.insert(path2It->first);
}
} else {
if (gDebugPrint)
vout << get(g_contigNames, pivot)
<< '\t' << path2 << '\n'
<< "ignored\t" << consensus << '\n';
overlaps.insert(id);
overlaps.insert(path2It->first);
}
}
cout << vout.str();
return id;
}
/** Return the length of the specified contig in k-mer. */
static unsigned getLength(const Lengths& lengths,
const ContigNode& u)
{
return u.ambiguous() ? u.length()
: lengths.at(u.id());
}
